Estimating distance to groups

[Tristen Brush]

Hi Eric, I'm doing CTDS on images (not stills from videos) of elk, who are (frustratingly, for me) VERY gregarious. I'm having issues with estimates coming out wildly different depending on how I treat group counts and distances, and I'm not sure how to proceed.

I've had reasonable success when using one record per image (rather than per individual), with the distance recorded to the individual that is nearest to the camera, and with the 'size' value equaling the group size as defined by the entire series of photos that photo belongs to (i.e., this number includes individuals that are not present in the current photo but are present in consecutive photos).

Reading further into the literature, I realized that it's suggested to measure distance either to every individual or to the average value of the group, rather than to the nearest individual. I also realized it doesn't make sense to use series-level group sizes if records represent individual photos. However, now I'm trying the models with image-level group sizes and to with distances to the average or every individual, and my estimates are way lower than expected. 

Do you have any advice for how to proceed? I understand logically why bringing down the group sizes is making estimates decrease, but is there something I might be overlooking that would make my estimates more reasonable without reverting to methods that don't make sense?

Thanks in advance,

Tristen